القائمة
Koru regression test harness workflow. Use when checking test status, investigating failures, or coordinating test runs with the user.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
Tend the membrane memory corpus — an OKF concept store whose fragment evolution lives in the git log as lineage trailers. Fire when creating, evolving, merging, splitting, or CORRECTING durable memory/knowledge that lives as OKF files under git; when recording how a belief changed or was repudiated; or when tracing a concept's history through time. YOU are the evolver — there is no evolution engine and no LLM sidecar. NOT for transient notes, scratch, or code comments.
Compile-time metaprogramming in Koru: event globbing and compile-time event matching — the "you are the compiler" surface where userland code sits inside the pipeline. Use when writing compile-time event selection or pattern-matching over events. SEED TOPIC — currently scoped to event globbing; widen `clusters` in koru-by-example.json as the metaprogramming surface grows.
Control flow in Koru: `~if` and `~for` as templates over continuation-return, not host keywords. Use when emitting conditionals or iteration in Koru, or when working on the template/lowering layer (`koru_std/control.kz`). Covers how `for`/`if` select continuations and lower per target.
How to work with the Koru language and its compiler `koruc`. Use when working in a Koru repo (`.kz`/`.k` files), running or reading regression tests, or when you need to orient on the toolchain — what the language is called, how to compile, where the docs and tests live. Start here before reading code.
How to work with the Koru language and its compiler `koruc`. Use when working in a Koru repo (`.kz`/`.k` files), running or reading regression tests, or when you need to orient on the toolchain — what the language is called, how to compile, where the docs and tests live. Start here before reading code.
Drive koru's diagnostic-code registry to green. Use when `zig run scripts/registry_check.zig` FIRES on DEAD/ORPHAN_EMIT/ROTTEN_PIN codes and you need to disposition each one (wire an emit / reserve / consolidate). The watcher detects drift in src/errors.zig's ErrorCode enum; this is the playbook for clearing it. Derived from three cold-room probes — trust the procedure, verify every claim.
| name | regression |
| description | Koru regression test harness workflow. Use when checking test status, investigating failures, or coordinating test runs with the user. |
The regression suite is the hub for all Koru compiler progress. ~600 tests, 40+ minutes for a full run.
The user runs the full suite (it takes too long for me to run):
./run_regression.sh # Full run
./run_regression.sh --parallel 8 # Parallel (faster)
I check what broke and investigate:
./run_regression.sh --regressions # What's failing + when it last passed
./run_regression.sh --status # Current state from disk markers
./run_regression.sh --diff # Compare current vs last snapshot
We debug specific tests together:
./run_regression.sh 501 # Run single test (by number)
./run_regression.sh 330 # Run range 330-339
./run_regression.sh --history 501 # When did this test break?
After fixing, verify:
./run_regression.sh 501 502 503 # Re-run affected tests
./run_regression.sh smoke # Quick sanity check
| Command | Purpose |
|---|---|
--status | Show current test state from disk markers |
--regressions | List failing tests + when they last passed |
--diff | Compare current run vs last snapshot |
--history <id> | Show test history across all snapshots |
--list | List all tests with descriptions |
--priority | Show tests marked as PRIORITY |
--last-run | Show results from last full run |
Tests are marked with files in their directory:
SUCCESS / FAILURE - Result of last runTODO - Not yet implementedSKIP - Intentionally skippedBROKEN - Known broken, needs fixBENCHMARK - Performance test, not run normallyPRIORITY - Urgent, needs attentionAfter each full run, a snapshot is saved to test-results/ with:
--diff and --history--regressions first when the user reports failures.